Evaluations at the Cluster Level

Abstract

In Chap. 1, we discussed policy analysis involving the statistical comparison of performance measures across experimental groups. Statistical comparison determines whether an observed difference between groups should be interpreted as an effect of the intervention. Guided by the study question, evaluation of a policy may be performed at the level of the cluster or the individual, as discussed in Sect. 1.5 [140]. With the cluster as the unit of inference, we introduced two performance measures in Chap. 8. Evaluations at the cluster level, which typically assess whether a performance criterion is met, may involve comparing proportions. For example, this approach would be appropriate in determining whether booking surgery in advance reduces the proportion of hospitals that have surgical delays for high-priority elective procedures, relative to some other method of booking the procedures. Alternatively, evaluations at the cluster level may involve comparing averages. For example, comparisons of averages would be appropriate for studying whether use of a pooled list, rather than individual lists, for surgical consultation appointments decreases the average clearance time for the appointments lists.